I wondered if it may be good to compare bebop in your dataset. It supports typescript/rust/etc. and performed nearly as well as avro in this benchmark: https://adamfaulkner.github.io/binary_formats_are_better_than_json_in_browsers.html
They also mentioned your results and suggested that using data with more strings may be helpful in providing a fair benchmark. Maybe worth considering having a string and number dataset? Maybe even considering with/without zlib like this rust-based benchmark: https://david.kolo.ski/rust_serialization_benchmark/
I acknowledge that each of those would double the amount of testing to do so I'm suggesting adding adding a 9th format and then 4x the variants so 4.5x the amount of work. Thank you for making this and I'm glad it exists as-is. 😅
I wondered if it may be good to compare bebop in your dataset. It supports typescript/rust/etc. and performed nearly as well as avro in this benchmark: https://adamfaulkner.github.io/binary_formats_are_better_than_json_in_browsers.html
They also mentioned your results and suggested that using data with more strings may be helpful in providing a fair benchmark. Maybe worth considering having a string and number dataset? Maybe even considering with/without zlib like this rust-based benchmark: https://david.kolo.ski/rust_serialization_benchmark/
I acknowledge that each of those would double the amount of testing to do so I'm suggesting adding adding a 9th format and then 4x the variants so 4.5x the amount of work. Thank you for making this and I'm glad it exists as-is. 😅